XSF Discussion - 2022-05-15

Random question -- should the example here have a `to` attribute? https://xmpp.org/extensions/xep-0308.html#example-1

deuill: yes, and example 2 should have a from that makes sense and is equal. But that said, with Carbons and MAM you can't rely on the receiving client actually supporting the right XEP

Oh, it's actually mentioned after the examples

(the examples might've been copied from a XEP where servers were queried from the user's account, in which case to/from may be omitted in the request/response)

I'll open a PR!

Just need to get my Shakespeare on

Yeah I wish we stopped this already

No need for more bourgeois literature in specs

really

Yes really

haha

I'm more concerned about the liberal use of different TLDs really

I see `.lit`, `.example`, `.net`

well, `.example` is at least reserved for such a purpose

> deuill: > 2022-05-15 05:36 (GMT+05:30) > Just need to get my Shakespeare on I believeth that the example of thou XEP shalt be more enlightening if it is attributeth by `to`. Hence I have submitted this request in hopes that thou may accepteth the changes of mine. Yes I am fun at parties ✏

Hark! I beseech thee, prithee accepteth mine changes, made true in thine light!

*plays darts*

perfection

Shakespeare is bourgeois? Someone has never read Shakespeare…

Shakespeare is a racist and misogynistic criminal who should be canceled.

let's build a time machine!

Well without the canceling part, I guess this will apply to basically anybody from this moment in time

let's cancel all of history then?

Giant Comet 2022!

Yeahhh ... Let build a time machine and kill all our past because they didnt fit modern standars

Don't need to. Tchaikovsky is already canceled.

> Shakespeare is a racist and misogynistic criminal who should be canceled. Calm down, I was just being ironic. 😃

lol

This seems to be heading in a very off-topic direction.

If you want to cancel your history, just clear your MAM database

edhelas, there's no way to do that!

edhelas, yay you got me an new argument for ephemeral messages!!

edhelas, 🏆️

s/argument/point/

What is a MAM database?

Message archive

Used for synchronization of clients that have been offline for some time.

Zash: Thanks.

Current best(?) practice is to keep it for about a week so all your phones can catch up after going trough looooooooong tunnels.

So "yay" I guess it's great that nobody wants to discuss this and dismisses the subject as yet another fantasy of woke cancelers? Ah sorry it's offtopic anyway.

I'm ready to discuss any subject. I don't have taboos!

is privacy the main constraint y u cant search mam on muc server as soon as you are in that room, no scrolling?

qwestion: you can query the MAM of a public room even without joining, you just need a client that will implement it

like no dlding, just as if searching googles index doesnt require to dld the "entire" internet first

Ge0rG: which cl?

qwestion: no. the main constraints are performance (most clients fetch one "page" or about 50 messages on entering a room, not the entire history) and client DB design

qwestion: the client you haven't written yet

get on it

wgreenhouse: just need to finish self-building all my personal manufactures first.../s

I share the desire; given that my usecase is largely MUCs and IRC channels, in principle I really could mostly settle for whatever is on MAM, and needing to carry around a client db that I manage to get into an unrecoverable state every several months just to have decent search kinda sucks.

wgreenhouse: y need to dld anything? the muc server could have an index, i send my search term, and dld msgs i.e. 10 at a time , a bit like serp pg1,

Converse.js can be configured to show history before joining, and can scroll up and load earlier history

Full text search is less common however.

qwestion: server MAM implementations have an index, just not one optimized for full text search, rather for requesting date ranges.

qwestion: so you'd just need to write a client that acts as a search crwawler and makes its own index

it doesn't necessarily need to keep fulltext around forever to do that

> Full text search is less common however. Zash: as in inexistent?

Zash: which cl

wgreenhouse: y not have prosody or ejabberd do this

qwestion: chicken and egg--it's of no benefit for servers to waste compute and disk indexing this way if no clients would query it

I'm talking about servers. Prosody doesn't have FTS, don't think ejabberd does. Maybe Openfire has?

There are clients with this, like Conversations, Dino (which I just used to search for "full text search" here)

But they only search their local archives, not the server

yep. Conversations is how I generally do this (hence my comment about how silly it feels carrying around a local db that I manage to corrupt every few months just to have decent search)

From where I sit, there's generally more focus o n the private chat use case, which means E2EE, which *really* complicates server-side full text search.

wgreenhouse: i think both s and cl can index, but cl doesnt need to to search i.e. google, while server needs to generate index, whether the client chooses to 1) dld serp pg1 equivalent 2)dld all content in index fomat or 3) dld some number of msgs to then generate index client side?

qwestion: I could not parse most of that :) the current use of MAM, for most clients that use it, is that the client calculates a time interval of which messages it's missing, and uses that to query the server for a date range. as well as perhaps implementing the "scroll up for more" behavior. so server implementers respond to that use case, and index by date. there is no call for them to build an FTS index if nobody queries it.

Zash: im focused on mucs, though omemo/e2ee can simply trigger a modal warning if client tries to fts serverside (method 1 or 2, above) ✏

> From where I sit, there's generally more focus o n the private chat use case, which means E2EE, which *really* complicates server-side full text search. Zash: yup. and that probably looks like, if anything, servers getting dumber and not smarter, less helpful for users primarily interested in public MUCs

Ah, the search tells me ejabberd does have some old non-standard `withtext` thing.

oo

and this quote: > Full text search is a whole problem domain > It requires extra database support, and often language-specific indexes, and so on -- MattJ

sounds right.

new xeps...i would value this new way of dlding index or just the fts results to my query for ie cats (thus sending all msgs containing the word cats, previous 10 msgs at a time startest with most recent, say, for each time i scroll to top)

and a problem domain that's getting less attention if most are aiming at competing with Signal and WhatsApp

https://xmpp.org/extensions/xep-0431.html

> extending their existing extensions to conform to the extension in this specification that's a hell of a piece of drafting :D

Praise be Dave

Openfire supports full text search, and should even allow for some fuzzy search

Zash: that xep defines which number(s)

?

> while server needs to generate index, whether the client chooses to 1) dld serp pg1 equivalent 2)dld all content in index fomat or 3) dld some number of msgs to then generate index client side

Zash:

I don't understand any of that.

XEP-0431 defines a way to discover whether the server supports FTS and the field in the search form to use

for _server-side_ FTS

If you want to do client-side FTS then just use XEP-0313 to get all the messages and do whatever you want with whatever your local database can do.

> XEP-0431 > Version 0.2.0 (2020-01-29)¶ > No More Beer 😀🤔

k thx

Hello

Hello, Hello

coleman Hi

coleman what

whats your name W